[ovirt-users] Fwd: Re: adding gluster domains

[Sahina Bose]

[+ Nir, ovirt-users]

Nir,

Do you know of possible reasons for vdsm error while creating metadata 
file? Operations from the mount point seem to work.

thanks
sahina


On 10/21/2015 05:37 AM, Brett Stevens wrote:
> Hi Sahina.
>
> Thanks for helping.
>
> 1. Tried the "optimize-for-virt" got a few info messages as expected.
> 2. tried new domain. I still get Error while executing action 
> AddGlusterFsStorageDomain: Error creating a storage domain's metadata.
>
> If I  manually mount the volume (mount -tglusterfs 
> sjcstorage04:/vmstore01 /tmp/mnt/) and check, I find that it mounts fine.
> looking into the mount it seems to create the volume metadata ( or 
> partially, at least) as  I see the domain dir. in this case, 
> __DIRECT_IO_TEST__e2e1b4eb-abbe-4427-887c-ee8f4d3f9060.
>
> I just dont get it, this has been across many builds and tests. If 
> everyone else can mount gluster. I must be doing something 
> fundamentally wrong, though I don't know what,
>
> thanks
>
> Brett Stevens
>
> On Tue, Oct 20, 2015 at 4:38 PM, Sahina Bose <sabose at redhat.com 
> <mailto:sabose at redhat.com>> wrote:
>
>
>
>     On 10/20/2015 10:13 AM, Ravishankar N wrote:
>>     Hi Sahina,
>>
>>     Would you have any pointers on this? Brett is not able to attach
>>     gluster replicate volumes for storage using ovirt (distribute
>>     works fine though). Would VDSM logs help? Step-15 below has the
>>     error he is getting.
>>
>>     Regards,
>>     Ravi
>>
>>     On 10/20/2015 09:14 AM, Brett Stevens wrote:
>>>     Hi Ravi.
>>>
>>>     Tried again with the latest 3.6 release candidate. I still have
>>>     the exact same problems. cannot add a glusterfs domain, (this
>>>     happens on simple 2 brick replicas as well as our desired 3
>>>     replica 1 arbiter volumes and distribute seems ok) yet if I ssh
>>>     into the given host and issue the command, mount -t glusterfs
>>>     ..... it mounts fine. Its also worth noting that I can create
>>>     simple replica gluster
>>>     I can provide logs again if you like, but I thought that given
>>>     I'm the only one with this issue, it would be worth checking on
>>>     my workflow first, I must be missing a step somewhere.
>>>
>>>     The following steps are assuming Im building a simple 2 brick
>>>     replica using tcp
>>>
>>>     1 build generic host from centos 7 repo.
>>>     2 add ovirt36 repo rpm
>>>     3 install ovirt-engine-setup
>>>     4 yum update, reboot
>>>     5 selinux to disabled (would like to set this up later, but
>>>     removed for clarity)
>>>     6 firewalld off
>>>     7 engine setup
>>>     8 login to console , change dc name and cluster name
>>>     9 add hosts to a virt only compute cluster ( i switch off auto
>>>     firewall config for simplicity at the moment)
>>>     10 create a gluster cluster and add hosts
>>>     11  assume networking is all fine so far its the simple
>>>     ovirtmgmt and a single storage net all hosts are pingable on all
>>>     interfaces
>>>     12 create brick dirs in /export/testshare/brick01 on both hosts
>>>     13 chown -R 36:36 /export/ and check ls -lad
>>>     14 from gui create volume use all defaults just add brick paths
>>>     to gui ok, creates volume no problems, use gui to start volume
>>>     (also have used gluster cli from host consoles)
>
>     Please click on  "Optimize for virt store" on the created volume
>     (calling this even after starting the volume should be ok)
>     This is the only obvious missing step - but I'm thrown off by the
>     fact that distribute volume works for you.
>
>>>     15 go to storage / add new domain. select gluserfs, name it,
>>>     take the default "use host" (its always the first in the compute
>>>     cluster as you would expect) leave all other defaults as is.
>>>     click ok and get "Error while executing action Add Storage
>>>     Connection: Internal Engine Error"
>>>     16 check mount by ssh to the "use host" login and run mount -t
>>>     glusterfs host:/volume /rhvs-......../mnt and it will mount fine.
>
>     Is SELinux disabled on all the gluster nodes as well? Is firewalld
>     disabled on all gluster nodes too?
>
>
>>>
>>>     again, the same result if I use an arbiter brick, or not. The
>>>     only gluster volume that works for me is distribute.
>>>
>>>     to recap, I have the following status
>>>
>>>     can manually mount glusterfs replica and distribute type volumes
>>>     on vhost server
>>>     can create domain using nfs (but backed by gluster volume) both
>>>     replica and distribute type volumes via ovirt gui
>>>     cab create domain using distribute type gluster volume in ovirt gui
>>>     cannot create domain using 2 node replicate type gluster volume
>>>     in ovirt gui
>>>     cannot create domain using4 node replicate type gluster volume
>>>     in ovirt gui
>>>     cannot create domain using 3 node replicate  with arbiter 1 type
>>>     gluster volume in ovirt gui
>>>
>>>
>>>     could this be a version mismatch problem? I'm using only centos
>>>     repo and ovirt repos to build these hosts
>>>
>>>
>>>     thanks
>>>
>>>     Brett Stevens
>>>
>>
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ovirt.org/pipermail/users/attachments/20151021/81eee883/attachment-0001.html>